On the Bioeconomics of Marine Reserves when Dispersal Evolves · 2019. 8. 13. · 18 1 Introduction...
Transcript of On the Bioeconomics of Marine Reserves when Dispersal Evolves · 2019. 8. 13. · 18 1 Introduction...
On the Bioeconomics of Marine Reserves when Dispersal Evolves
Emily A. Moberg
Biology Department, MS #34, Woods Hole Oceanographic Institution, Woods Hole, MA 02543
Esther Shyu
Biology Department, MS #34, Woods Hole Oceanographic Institution, Woods Hole, MA 02543
Guillermo E. Herrera
Department of Economics, Bowdoin College, 9700 College Station, Brunswick, ME 04011
Suzanne Lenhart
Mathematics Department, University of Tennessee, Knoxville, TN 37996-1300
Yuan Lou
Institute for Mathematical Sciences, Renmin University of China, Beijing, 100872, PRC and
Department of Mathematics, Ohio State University, 231 W 18th Ave, MW 450, Columbus, OH 43210
Michael G. Neubert∗
Biology Department, MS #34, Woods Hole Oceanographic Institution, Woods Hole, MA 02543
Published November 2015 in: Natural Resource Modeling 28 (4):456-474.
∗Corresponding author: [email protected]
Abstract1
Marine reserves are an increasingly used and potentially contentious tool in fisheries manage-2
ment. Depending upon the way that individuals move, no-take marine reserves can be necessary for3
maximizing equilibrium rent in some simple mathematical models. The implementation of no-take4
marine reserves often generates a redistribution of fishing effort in space. This redistribution of5
effort, in turn, produces sharp spatial gradients in mortality rates for the targeted stock. Using6
a two-patch model, we show that the existence of such gradients is a sufficient condition for the7
evolution of an evolutionarily stable conditional dispersal strategy. Thus, the dispersal strategy8
of the fish depends upon the harvesting strategy of the manager and vice versa. We find that an9
evolutionarily stable optimal harvesting strategy (ESOHS)—one which maximizes equilibrium rent10
given that fish disperse in an evolutionarily stable manner—never includes a no-take marine re-11
serve. This strategy is economically unstable in the short run because a manager can generate more12
rent by disregarding the possibility of dispersal evolution. Simulations of a stochastic evolutionary13
process suggest that such a short-run, myopic strategy performs poorly compared to the ESOHS14
over the long run, however, as it generates rent that is lower on average and higher in variability.15
Keywords: evolution of dispersal, evolutionarily stable strategy, fisheries management, marine pro-16
tected areas, optimal harvesting.17
1 Introduction18
No-take marine reserves are a type of “marine protected area” in which fishing is prohibited. Closed19
areas like marine reserves have been used to manage artisanal fisheries on small spatial scales for20
many years (Fogarty et al., 2000). The advent of geographical positioning systems (which make21
the possibility of enforcing closures more feasible (Pala, 2014)) combined with the decline of fish22
stocks, an increased demand for marine fish protein (FAO Fisheries Department, 2014), and a call23
for ecosystem-based management, have led not only to increased study of the efficacy of marine24
reserves but also to an increase in their implementation. Marine protected area coverage worldwide25
has increased by over 150% since 2003 (Toropova et al., 2010).26
A number of studies have shown that marine reserves can contribute to the conservation of27
stocks and to the ecosystems that support them (e. g., Halpern and Warner, 2002; Halpern, 2003;28
Lester et al., 2009). Increases in individual size, biomass, population density and species diversity29
have been shown to increase subsequent to reserve establishment (see examples in, for example,30
Lester and Halpern, 2008).31
The potential economic costs or benefits of reserves are less clear (Kaiser, 2005; White et al.,32
2008; Hart and Sissenwine, 2009; Fletcher et al., 2015, in press). Some modeling studies (e. g.,33
Neubert, 2003; Sanchirico and Wilen, 2005; Sanchirico et al., 2006; Armstrong, 2007; Neubert and34
Herrera, 2008; Joshi et al., 2009; Moeller and Neubert, 2013) have shown that the establishment35
of marine reserves for conservation purposes does not necessarily require a reduction in economic36
productivity. Indeed, in some models reserves are necessary to maximize yield or sustainable rent.37
Others (including Polacheck, 1990; Quinn et al., 1993; Man et al., 1995; Holland and Brazee, 1996;38
Nowlis and Roberts, 1999; Guenette and Pitcher, 1999; Hastings and Botsford, 1999; Li, 2000;39
Pezzey et al., 2000; Sanchirico and Wilen, 2001; Apostolaki et al., 2002) have shown that reserves40
may be yield-neutral or produce minor improvements when compared with non spatial effort-control41
policies. In some cases, the establishment of a reserve decreases yield (Tuck and Possingham, 1994).42
The optimality of reserves, then, would seem to depend both on the objective as well as the43
ecological and economic circumstances. One phenomena, however, emerges from all of these mod-44
eling studies, as well as from real-world observations (Fig. 1): the imposition of marine reserves45
can produce a radical redistribution of fishing effort in space. Effort is displaced from reserve areas46
1
and frequently concentrates near their borders as harvesters attempt to catch the “spillover” from47
the reserves. As a consequence, the establishment of marine reserves can produce sharp spatial48
gradients in mortality (Neubert, 2003; Kellner et al., 2007; Joshi et al., 2009; Abbott and Haynie,49
2012; Moeller and Neubert, 2013).50
It is easy to imagine, that as a result of these gradients, there would be strong selective pressure51
to evolve context-dependent dispersal (McPeek and Holt, 1992)—that is, low dispersal rates within52
the reserve and high dispersal rates outside—or, equivalently, the ability for dispersing individuals53
to detect and preferentially settle in better patches. Since the potential economic benefits of reserves54
rely on dispersal of individuals from reserves into fished areas, evolution of dispersal might work55
against the generation of sustainable rent.56
In this paper we explore that possibility with the aid of a simple,“two-patch” model (Holt, 1985).57
We begin by briefly demonstrating that, in the absence of evolution, reserves can be economically58
optimal when the two patches are sufficiently different in either their biological or economic prop-59
erties (Sanchirico et al., 2006). We then ask whether reserves are ever optimal (in the sense of60
maximizing equilibrium rent) when dispersal evolves.61
Our analysis of this second problem builds on the work of Law and Grey (1989) and Grey62
(1993) who were perhaps the first to seriously investigate the interplay between harvest and evo-63
lution, i. e., the inclusion of evolutionary change in the constrained optimization problem of the64
resource manager. They developed the concept of an evolutionarily stable optimal harvest strat-65
egy (ESOHS)1—a harvesting strategy “which gives the greatest sustainable yield, after evolution66
caused by cropping has taken place.” Law and Grey (1989) were particularly concerned with the67
problem of how age-specific harvesting selects for changes in the age at maturity, so they developed68
the ESOHS concept in the context of life-history theory (which generally ignores dispersal). We69
extend their idea here to the evolution of dispersal in a spatially managed fishery and find that70
evolution qualitatively changes the nature of the optimal distribution of fishing effort.71
1We prefer the pronunciation ess-oh-ess for this acronym.
2
2 Model72
The model we use is similar to those of Clark (1990, pg. 337) and Sanchirico et al. (2006), both73
of which derive from the classic model of Gordon (1954). The model describes the dynamics of a74
stock distributed across two spatial locations, or “patches,” connected by dispersal. Each patch is75
characterized by an intrinsic rate of growth ri and a carrying capacity ki. Individuals leave a patch76
at a constant per capita rate m and enter a common pool of dispersers. From this pool a fraction77
ε (instantaneously) choose to settle into patch 1; the remaining fraction, 1 − ε, settle in patch 2.78
In this sense, ε can be thought of as a disperser’s preference for patch 1. Patches are harvested at79
nonnegative patch-dependent effort rates Ei. If the population size of the stock in patch i is xi,80
this fishing effort generates yield at the rate qiEixi. The proportionality constants qi are called the81
“catchability coefficients.”82
Under this model, the dynamics of the stock in the two patches are given by the ordinary83
differential equations84
dx1dt
= r1x1
(1 − x1
k1
)−m(1 − ε)x1 +mεx2 − q1E1x1, (1)
dx2dt
= r2x2
(1 − x2
k2
)+m(1 − ε)x1 −mεx2 − q2E2x2. (2)
If the price of fish is p, and the cost per unit of effort in patch i is ci, then the rent generated85
by harvesting is86
π[E1, E2; ε] =
2∑i=1
(pqixi − ci)Ei. (3)
At first, we concern ourselves with the case in which a manager is able to control the levels of effort87
in each of the patches (for example by limiting the number of boat-days available for fishing or by88
taxing effort) and does so with the objective of maximizing the rent, π, at equilibrium.89
It is a simple matter to numerically calculate the equilibrium stock sizes from equations (1) and90
(2) for any combination of E1 and E2. These can be substituted into formula (3) to determine the91
equilibrium rent. We call the effort levels that maximize the equilibrium rent E∗i , the corresponding92
stock sizes x∗i , and the maximum equilibrium rent π∗.93
The optimal solution in patch i will fall into one of three categories depending upon the signs94
3
of E∗i and the marginal rent in patch i, pqix∗i − ci. If95
1. E∗i > 0, we say the patch is fished ; if96
2. E∗i = 0 and pqix∗i − ci ≤ 0, we say the patch is unfished ; and if97
3. E∗i = 0 and pqix∗i − ci > 0, we say the patch is in reserve.98
We distinguish between unfished and reserve patches because the latter would require enforcement99
by the regulator—an individual harvester would have incentive to fish in that patch, but doing so100
would reduce the total rent at equilibrium. In unfished patches the marginal rent is negative, and101
rational harvesters (which we assume) avoid it of their own accord.102
The optimal equilibrium effort levels in each patch are determined by the model parameters103
(Fig. 2). When the patches are economically and ecologically identical, and dispersers settle indif-104
ferently (i. e., ε = 0.5), the optimal strategy is to ensure that both patches are harvested at the105
same rate (or not fished at all if pqiki − ci ≤ 0). Asymmetric settlement, or differences in intrinsic106
growth rates, carrying capacities, or harvest costs can result in the optimal closing of one patch107
(blue and red regions of Fig. 2). For the rest of the paper, we will explore cases in which patch108
1 is in one way (and only one way) better (for the harvesters) than patch 2; that is, all of the109
inequalities110
r1 ≥ r2, k1 ≥ k2, c1 ≤ c2, q1 ≥ q2, (4)
are satisfied and only one is satisfied as a strict inequality. This is the case for all of the parameter111
combinations encompassed by Fig. 2 and subsequent figures.112
3 Evolution of dispersal and the ESS113
In general, the optimal harvesting effort, and thus the per capita mortality rate, in each patch114
will differ. The dispersal strategy may evolve in response to this mortality gradient. Evolution,115
in turn, affects optimal fishing strategies, including the optimality of reserves, through changes in116
dispersal. Here, we consider the evolution of ε, the proportion of dispersers that settle into patch117
1. We derive the evolutionarily stable strategy (ESS), ε, the dispersal phenotype against which no118
alternative phenotype can increase under selection. In this section, we find an expression for ε and119
4
show that it is a “weak form ESS.” This ESS is also convergence-stable, making it an evolutionary120
attractor to which the population will converge in the long run.121
3.1 Calculating the ESS122
To determine ε, we begin by considering a population composed of a single “resident” phenotype
with dispersal preference ε. The equilibrium stock sizes, x1 and x2, satisfy
[r1
(1 − x1
k1
)− q1E1
]x1 −m(1 − ε)x1 +mεx2 = 0, (5)[
r2
(1 − x2
k2
)− q2E2
]x2 +m(1 − ε)x1 −mεx2 = 0. (6)
We will find it useful to define αi as the per capita growth rate, including fishing mortality, in patch123
i if it were isolated (i. e., if m = 0). That is,124
αi =
[ri
(1 − xi
ki
)− qiEi
]. (7)
αi can be thought of as the fitness of an individual in patch i at equilibrium.125
The phenotype that characterizes the resident population evolves through invasions (and se-126
quential replacement) by rare mutants—alternative phenotypes that appear at low frequencies.127
Mutants are identical to residents, save for their dispersal preference, which we will denote as ε′.128
A mutant’s fate depends on its invasion fitness—its initial growth rate in the resident population.129
When it first appears, the mutant is rare, and its effect on the resident’s population dynamics is130
negligible (Metz, 2008). Thus if x′1 and x′2 are the mutant populations in the two patches, their131
dynamics are initially given by the linear system132
d
dt
x′1
x′2
= A′
x′1
x′2
(8)
where133
A′ =
α1 −m(1 − ε′) mε′
m(1 − ε′) α2 −mε′
. (9)
5
The invasion fitness is then given by the dominant eigenvalue of A′ (which is always real):134
λ′ =1
2
(α1 + α2 −m+
√(α1 − α2)2 + 2(α1 − α2)(2ε′ − 1)m+m2
). (10)
Note that the invasion fitness is a function of both the mutant phenotype and the resident phenotype135
(because the α’s depend upon the equilibrium population sizes of the resident, which, in turn depend136
on ε).137
If the invasion fitness (10) is positive, the mutant can replace the resident, inducing evolutionary138
change; if negative, the mutant will be extirpated. An ESS, ε, is a resident phenotype that cannot139
be replaced by any ε′, making it resistant to further evolution (Geritz et al., 1998). A condition140
that must be satisfied by any ESS is that the selection gradient dλ′/dε′ vanishes when ε′ = ε = ε.141
Differentiating the invasion fitness (10) with respect to ε′ and evaluating at ε′ = ε = ε gives142
∂λ′
∂ε′
∣∣∣∣ε′=ε=ε
=(α1 − α2)m√
(α1 − α2)2 + 2(α1 − α2)(2ε− 1) +m2= 0. (11)
Since we have assumed that m is positive, a vanishing selection gradient (11) implies that143
α1 = α2; but, adding (5) and (6) we find that144
α1x1 + α2x2 = α1(x1 + x2) = 0. (12)
Thus, when the resident population sizes are positive, α1 = α2 = 0. That is, when the patch145
preference is at its ESS value, ε, the per capita growth rates in the two patches (including fishing146
mortality) are identical and zero.147
By setting α1 = α2 = 0 in equilibrium equations (5) and (6), we see that the only potential148
ESS is149
ε =x1
x1 + x2, (13)
where150
xi = ki
(1 − qiEi
ri
)(14)
are the corresponding population sizes.151
Substituting the condition α1 = α2 = 0 into (10), we see that the invasion fitness of any mutant152
6
is 0 whenever the resident phenotype is given by (13). Because the invasion fitness is never positive,153
no mutant phenotype can increase under selection, confirming that (13) is a local ESS. Because154
the invasion fitness is always 0, however, every mutant will have the same fitness as the resident,155
making (13) a ‘weak form ESS’ (sensu Uyenoyama and Bengtsson, 1982).156
3.2 Convergence stability of the ESS157
As we show next, the evolutionarily stable dispersal strategy (13) is also convergence stable—an158
evolutionary attractor to which a monomorphic population will converge through small, successive159
mutations (Geritz et al., 1998). We thus expect the settlement preference to evolve to, and remain160
at, ε.161
We demonstrate the convergence stability of ε using the second derivatives of the invasion fitness162
(10). Convergence stability requires that163
(∂2λ′
∂ε ∂ε′+∂2λ′
∂ε′2
) ∣∣∣∣ε′=ε=ε
< 0. (15)
That is, the sum of these second derivatives, taken with respect to the resident and mutant pheno-164
types, must be negative at the ESS ε (Eshel, 1983; Geritz et al., 1998).165
Because α1 and α2 do not depend on the mutant strategy ε′, it follows that ∂2λ′/∂ε′2 = 0 when166
α1 = α2. Thus, (13) will be a convergence-stable ESS if ∂2λ′/∂ε ∂ε′ < 0 at ε′ = ε = ε.167
To calculate ∂2λ′/∂ε∂ε′, first differentiate the invasion fitness (10) with respect to ε′:168
∂λ′
∂ε′=
(α1 − α2)m√(α1 − α2)2 + 2(α1 − α2)(2ε′ − 1)m+m2
. (16)
Next, recall that α1 and α2 depend on the resident trait ε, and rewrite the equilibrium conditions
(5) and (6) as
α1 = m
[1 − ε(x1 + x2)
x1
], (17)
α2 = m
[ε− (1 − ε)x1
x2
]. (18)
Note that the equilibrium stock sizes x1 and x2 are both functions of ε.169
7
We can substitute (17) and (18) into (16), and then differentiate with respect to ε to obtain170
∂2λ′/∂ε∂ε′. After evaluating the resulting expression at ε′ = ε = ε, as given by (13), we find that171
∂2λ′
∂ε ∂ε′
∣∣∣∣ε′=ε=ε
=m
x1x2
[x2dx1dε
− x1dx2dε
− (x1 + x2)2
]. (19)
The derivatives dx1/dε and dx2/dε can be found by differentiating the equilibrium equations
(5) and (6) with respect to ε. When evaluated at ε′ = ε = ε and xi = xi, as given by (14), these
derivatives are
dx1dε
∣∣∣∣ε′=ε=ε
=mk1r2x2(x1 + x2)
2
mk2r1x21 + r2x2 [mk1x2 + r1x1(x1 + x2)], (20)
dx2dε
∣∣∣∣ε′=ε=ε
= − mk2r1x1(x1 + x2)2
mk2r1x21 + r2x2 [mk1x2 + r1x1(x1 + x2)]. (21)
After substituting (20) and (21) into (19), we find that172
∂2λ′
∂ε ∂ε′
∣∣∣∣ε′=ε=ε
= − m
x1x2
(r1r2x1x2(x1 + x2)
3
mk2r1x21 + r2x2 [mk1x2 + r1x1(x1 + x2)]
)< 0. (22)
It follows that inequality (15) is satisfied and the ESS settlement preference (13) is a convergence-173
stable strategy.174
4 The ESOHS and effects of evolution on optimal management175
In general, the rent that is generated in each patch depends upon the fishing effort in both patches.176
This is not the case when the patch preference ε is at its ESS value ε, which becomes clear upon177
substituting the equilibrium stock sizes (14) into the rent (3):178
π[E1, E2; ε] = π =
2∑i=1
(pqiki
(1 − qiEi
ri
)− ci
)Ei. (23)
This means that when we maximize rent over E1 and E2, we are maximizing the rent in the patches179
independently of each other. Thus, a reserve cannot be part of an ESOHS; a patch should never180
be closed unless it is unprofitable to harvest (i. e., falls in the ‘unfished’ category). Specifically, the181
8
ESOHS is182
E∗i =
ri(pqiki−ci)
2pq2i kiif pqiki − ci > 0,
0 otherwise.(24)
The resulting stock sizes in each patch at the ESOHS are183
x∗i =
12
(ki + ci
pqi
)if pqiki − ci > 0,
ki otherwise.(25)
The evolutionarily stable settlement preference at optimal harvest, ε∗, can be calculated using (13)184
with stock sizes xi = x∗i .185
Spatial heterogeneity in biological or economic parameters is reflected in the ESOHS (Fig. 3).186
When the patches differ in their biological parameters (r or k), the ESOHS effort level in the187
worse patch is smaller than it would be if the patches were identical and the parameter values188
were equal to their values in the good patch (Fig. 3, first two columns). If the only difference189
between the patches is due to a difference in intrinsic growth rate (i. e., if r2 < r1), the ESOHS190
settlement preference, ε∗, remains 1/2, and the stock sizes are equal to one half of the (identical)191
carrying capacity in each patch. In contrast, when the carrying capacities of the two patches differ192
(i. e., k2 < k1), ε > 1/2, and settlement in patch 1 is more frequent than settlement in patch193
2. In combination with the lower carrying capacity, this dispersal asymmetry results in a smaller194
equilibrium stock size in patch 2.195
When the patches differ in one of their economic parameters (either c or q; Fig. 3, last two196
columns), ε∗ < 1/2; that is, settlement is more frequent in the economically poorer patch. If the197
patches only differ in the cost of fishing (i. e., c2 > c1), then the ESOHS effort in the more expensive198
patch, as expected, is lower than in the less expensive patch. Combined with the settlement199
asymmetry, this results in a larger standing stock in the poorer patch. Similarly, there is a larger200
standing stock in patch 2 when fish are harder to catch there (i. e., q2 < q1). In contrast with201
differences in cost, however, the ESOHS effort level in the patch with lower catchability (E∗2) is202
higher than it is in the patch where fish are easier to catch (at least until fish become so difficult203
to catch that it is no longer worth harvesting in patch 2 at all).204
9
4.1 Management with reserves205
Marine reserves may be part of an economically optimal, equilibrium management strategy when206
dispersal does not evolve; however, as (24) shows, this is not the case when dispersal does evolve.207
While marine reserves are not part of the ESOHS, they may be desirable for other purposes. It is208
therefore interesting to know how the establishment of a reserve would impact profits. The impact209
of a reserve is contingent upon whether the organisms evolve in response to differences in growth210
or mortality conditions.211
We placed either patch 1 or patch 2 in reserve and calculated the unconstrained rent-maximizing212
level of effort in the other patch. We also calculated the effort level when the resulting settlement213
preference was constrained to be evolutionarily stable. We found that using reserves when the214
settlement preference ε evolves can produce dramatically lower profits (Fig. 4). When a patch is215
placed in reserve, ε evolves to increase the tendency of fish to disperse to that patch (i.e., when216
patch 1 is in reserve, ε increases relative to its value when both efforts are optimized to the ESS217
settlement preference). At least for the parameter values we studied, ε varies most with variation218
in k2 and varies least with r2 (Fig. 4, top row).219
4.2 Is the ESOHS economically stable?220
The ESOHS represents the best equilibrium harvesting strategy under the constraint that the221
strategy will not produce further evolutionary change. At the ESOHS no mutant phenotypes can222
invade and displace the resident phenotype. We have assumed that those mutants are rare, so that223
there will generally be a long time between mutation events. In between such events, however, the224
ESOHS is suboptimal. More rent could be extracted from the resource if the manager were to set225
the effort levels at their unconstrained levels (i. e., π[E∗1 , E∗2 ; ε∗] ≤ π[E∗1 , E
∗2 ; ε∗]), and the manager226
will be sorely tempted to do so. As a result, we should not expect the ESOHS to be economically227
stable.228
As a consequence of fishing at (short-term) optimal levels, rather than according to the ESOHS,229
the resident phenotype would no longer be an ESS and would be vulnerable to an invasion by a more230
fit mutant. Of course the manager could simply change his or her harvesting strategy to optimize231
the rent given this new phenotype. Because of the way it disperses, the potential profitability of232
10
a new phenotype would likely be different than that of the resident. Imagine that this iterative233
process—harvesting at rent-maximizing rates, invasion of a new phenotype, adjustment of the234
harvesting rates, etc.—continued for a long time. At some times the instantaneous rent would be235
larger than that that could be generated by the ESOHS; in some instances, it would be less.236
We simulated this “reactionary” policy by introducing a mutant phenotype according to a237
Poisson process with rate constant µ. We drew the mutant phenotype ε′ from a normal distribution238
with mean equal to the resident phenotype ε, and standard deviation σ, truncated so that 0 < ε′ <239
1. Whenever a mutant appeared, we computed the invasion fitness (10). If the invasion fitness240
was positive, we replaced the resident by the mutant phenotype and calculated a new harvesting241
policy that would maximize equilibrium rent for the new phenotype. (In doing so, we implicitly242
assume that invasion implies displacement. For sufficiently small mutations, Geritz et al. (2002)243
have proved that this substitution does occur.)244
We show a single realization of such a reactionary harvesting policy in Fig. 5. When the mutant245
invades, the efforts in each patch, the population levels, and the profits also fluctuate. In the case246
illustrated, ε tends to be less than the ESOHS ε value, while the effort and population levels tend247
to be higher than the ESOHS level in patch 1 (blue lines) and lower in patch 2 (orange lines). The248
rent derived from the reactionary policy tends to be less than the ESOHS rent for this realization.249
We simulated this stochastic process for a variety of parameter values to assess the average250
performance of a reactionary versus ESOHS harvesting policy; we found that the rent generated251
by the ESOHS always exceeded the average rent generated by reactionary harvesting (Fig. 6, top252
row). It appears that, on average, harvesting at rates that maximize short-term profits selects for253
new phenotypes that are inimical to expected long-term sustainable rent. In addition to boosting254
average rent, using the ESOHS has the additional advantage of reducing (to zero) the variability in255
profits that would accompany reactionary harvesting (Fig. 6, bottom row). Our simulations suggest256
that the more different the two patches are, the lower and the more variable are the reactionary257
rents.258
11
5 Discussion259
In a simple two-patch model, we have shown that almost every optimal harvesting strategy is260
unstable in the face of dispersal evolution. The exception is a unique evolutionarily stable optimal261
harvesting strategy, or ESOHS, where dispersal, as described by the settlement preference, is a262
convergence-stable, weak-form ESS. The ESOHS, however, is potentially economically unstable: in263
the short term, a manager could always generate more rent using a different distribution of effort264
(sometimes using a reserve), at least until a new phenotype invades. A manager who employs a265
myopic, reactionary strategy of constantly maximizing equilibrium rent, assuming that the current266
phenotype will not change, suffers reduced average rent, and higher variation in rent, over long267
timescales. In the real world, there would be economic and social benefits of a consistent harvest268
strategy, compared to one that changed unpredictably in response to evolutionary changes.269
Marine reserves do not play a role in the ESOHS for the two-patch model. This is because270
evolution of dispersal acts to equalize fitness between the two patches and push population densities271
to levels that result in no net movement between them. Without this net movement of individuals,272
or “spillover,” from the reserve patch into the fished patch, reserves only reduce economic benefits.273
The equilibration of fitness across habitats is the sine qua non of the so-called ideal free distribution274
(Fretwell and Lucas, 1969). Based on our results with the two-patch model, we conjecture that,275
more generally, marine reserves will never be economically optimal when the dispersal behavior of276
individuals leads to the ideal free distribution of the population. The evolution of dispersal, however,277
does not inevitably lead to the ideal free distribution. In particular, the ideal free distribution does278
not emerge as the result of an evolutionary stable dispersal strategy when the environment has a279
source-sink structure and is characterized by temporal variability in fitness (Holt and Barfield, 2001;280
Schreiber, 2012). Describing the ESOHS in such circumstances, if one exists, would be challenging.281
Our results, when combined with the results from Baskett et al. (2007), who found that in-282
creased fragmentation of a reserve network tended to reduce dispersal distance (i. e., increase local283
retention), suggests that evolution of dispersal may be an important consideration for spatially284
managed fisheries. However, our understanding of the likely effects of dispersal evolution on opti-285
mal management is still nascent. For example, dispersal may encompass a host of traits, including286
larval duration, the proportion of offspring which disperse or migrate (a la Baskett et al., 2007;287
12
Dunlop et al., 2009), or adaptive movements of mature individuals (a la Abrams et al., 2012). How288
reserves impact population sizes and selection pressures will depend on the particular dispersal289
trait.290
Of course, settlement preference is not the only life history trait that may evolve in response291
to harvesting (Borisov, 1978; Jørgensen et al., 2007; Allendorf et al., 2008; Heino and Dieckmann,292
2009). Most other studies have focused on size-selective harvest, evolution of age or size at maturity293
(Kuparinen and Merila, 2007) and the consequences (both negative and positive) that such fisheries294
induced evolution can have on sustainable yield or rent(Law and Grey, 1989; Heino, 1998; Law,295
2000; Ratner and Lande, 2001; Eikeset et al., 2013). Intriguingly, it has been suggested that296
marine reserves might ameliorate the consequences of fisheries induced evolution of such traits297
(Baskett et al., 2005; Miethe et al., 2010). The ramifications of marine reserves in real evolving298
systems are likely to be complicated by the simultaneous evolution of multiple traits which may299
have countervailing effects.300
While our study suggests that evolution of dispersal may reduce the efficacy of reserves as a rent-301
maximizing strategy, our analysis focused on equilibrium management on very long timescales. As302
Sanchirico et al. (2006) highlighted, solving for the optimal harvest trajectory between two patches303
through time is much more difficult; different results regarding marine reserve optimality may304
emerge in this case.305
Acknowledgements306
The authors acknowledge helpful discussions with S. Cantrell, C. Cosner, H. Caswell, and the307
participants in the workshop on “Rapid Evolution and Sustainability” at the Mathematical Bio-308
sciences Institute at Ohio State University. They are also thankful for the hospitality of the Banff309
International Research Station. This material is based upon work supported by funding from: The310
Woods Hole Oceanographic Institution’s Investment in Science Fund to MGN; The Recruitment311
Program of Global Experts to YL; The University of Tennessee Center for Business and Economics312
Research to SL; and the U.S. National Science Foundation (NSF) through grants OCE-1031256,313
DEB-1257545, and DEB-1145017 to MGN, CNH-0707961 to GEH, DMS-1411476 to YL; and NSF314
Graduate Research Fellowships under Grant No. 1122374 to EAM and ES.315
13
References316
Abbott, J. K. and Haynie, A. C. (2012). What are we protecting? fisher behavior and the unin-317
tended consequences of spatial closures as a fishery management tool. Ecological Applications,318
22(3):762–777.319
Abrams, P. A., Ruokolainen, L., Shuter, B. J., and McCann, K. S. (2012). Harvesting creates320
ecological traps: consequences of invisible mortality risks in predator-prey metacommunities.321
Ecology, 93(2):281–293.322
Allendorf, F. W., England, P. R., Luikart, G., Ritchie, P. A., and Ryman, N. (2008). Genetic effects323
of harvest on wild animal populations. Trends in Ecology & Evolution, 23(6):327–337.324
Apostolaki, P., Milner-Gulland, E. J., McAllister, M., and Kirkwood, G. (2002). Modelling the325
effects of establishing a marine reserve for mobile fish populations. Canadian Journal of Fisheries326
and Aquatic SciencesJ. Fish Aquat. Sci., 59:405–415.327
Armstrong, C. W. (2007). A note on the ecological-economic modelling of marine reserves in328
fisheries. Ecological Economics, 62:242–250.329
Baskett, M. L., Levin, S. A., Gaines, S. D., and Dushoff, J. (2005). Marine reserve design and the330
evolution of size at maturation in harvested fish. Ecological Applications, 15(3):882–901.331
Baskett, M. L., Weitz, J. S., and Levin, S. A. (2007). The evolution of dispersal in reserve networks.332
The American Naturalist, 170(1):59–78.333
Borisov, V. M. (1978). The selectivve effect of fishing on the population structure of species with334
long life cycle. Journal of Icthyology, 18:896–904.335
Clark, C. W. (1990). Mathematical Bioeconomics: The Optimal Management of Renewable Re-336
sources. John Wiley and Sons, New York.337
Dunlop, Erin, S., Baskett, M. L., Heino, M., and Dieckmann, U. (2009). Propensity of marine338
reserves to reduce the evolutionary effects of fishing in a migratory species. Evolutionary Appli-339
cations, 2:371–393.340
14
Eikeset, A. M., Richter, A., Dunlop, E. S., Dieckmann, U., and Stenseth, N. C. (2013). Economic341
repercussions of fisheries-induced evolution. Proceedings of the National Academy of Sciences of342
the USA, 110:12259–12264.343
Eshel, I. (1983). Evolutionary and continuous stability. Journal of Theoretical Biology, 103:99–111.344
FAO Fisheries Department (2014). State of world fisheries and aquaculture (sofia). Technical345
report, Food and Agriculture Organization of the United Nations.346
Fletcher, W. J., Kearney, R. E., Wise, B. S., and Nash, W. J. (2015). Large-scale expansion of347
no-take closures within the Great Barrier Reef has not enhanced fishery production. Ecological348
Applications, 25:1187–1196.349
Fogarty, M., Bohnsack, J., and Dayton, P. (2000). Marine reserves and fishery management. In350
Sheppard, C., editor, Seas at the Millennium: An Environmental Evaluation, pages 283–300.351
Elsevier.352
Fogarty, M. and Murawski, S. A. (2004). Do marine protected areas really work? Oceanus, 43(2).353
Fretwell, S. D. and Lucas, H. L. J. (1969). On territorial behavior and other factors influencing354
habitat distribution in birds. Acta Biotheoretica, 19:16–36.355
Geritz, S. A. H., Gyllenberg, M., Jacobs, F. J. A., and Parvinen, K. (2002). Invasion dynamics and356
attractor inheritance. Journal of Mathematical Biology, 44:548–560.357
Geritz, S. A. H., Kisdi, E., Meszena, G., and Metz, J. A. J. (1998). Evolutionarily singular strategies358
and the adaptive growth and branching of the evolutionary tree. Evolutionary Ecology, 12:35–57.359
Gordon, H. (1954). The economic theory of a common-property resource: the fishery. Journal of360
Political Economy, 62:124–142.361
Grey, D. R. (1993). Evolutionarily stable optimal harvesting strategies. In The Exploitation of362
Evolving Resources, pages 176–186. Springer-Verlag.363
Guenette, S. and Pitcher, T. J. (1999). An age-structured model showing the benefits of marine364
reserves in controlling overexploitation. Fisheries Research, 39:295–303.365
15
Halpern, B. S. (2003). The impact of marine reserves: do reserves work and does reserve size366
matter? Ecological Applications, 13:S117–S137.367
Halpern, B. S. and Warner, R. R. (2002). Marine reserves have rapid and lasting effects. Ecology368
Letters, 5:361–366.369
Hart, D. R. and Sissenwine, M. P. (2009). Marine reserve effects on fishery profits: a comment on370
white et al. (2008). Ecology Letters, 12:E9–E11.371
Hastings, A. and Botsford, L. W. (1999). Equivalence of yield from marine reserves and traditional372
fisheries management. Science, 284:1537–1538.373
Heino, M. (1998). Management of evolving fish stocks. Canadian Journal of Fisheries and Aquatic374
Sciences, 55:1971–1982.375
Heino, M. and Dieckmann, U. (2009). Fisheries-induced Evolution. Wiley Online Library.376
Holland, D. S. and Brazee, R. J. (1996). Marine reserves for fisheries management. Mar. Res.377
Econ., 11:157–171.378
Holt, R. D. (1985). Population dynamics in two-patch environments: some anomalous consequences379
of an optimal habitat distribution. Theoretical Population Biology, 28(2):181–208.380
Holt, R. D. and Barfield, M. (2001). On the relationship between the ideal free distribution and the381
evolution of dispersal. In Clobert, J., Danchin, E., Dhondt, A. A., and Nichols, J. D., editors,382
Dispersal, pages 83–95. Oxford University Press.383
Jørgensen, C., Enberg, K., Dunlop, E. S., Arlinghaus, R., Boukal, D. S., Brander, K., Ernande,384
B., Gardmark, A., Johnston, F., Matsumura, S., et al. (2007). Managing evolving fish stocks.385
Science, 318(5854):1247–1248.386
Joshi, H. R., Herrera, G. E., Lenhart, S., and Neubert, M. G. (2009). Optimal dynamic harvest of387
a mobile renewable resource. Natural Resource Modeling, 22(2):322–343.388
Kaiser, M. J. (2005). Are mainre protected areas a red herring or fisheries panacea? Canadian389
Journal of Fisheries and Aquatic Sciences, 62:1194–1199.390
16
Kellner, J. B., Tetreault, I., Gaines, S. D., and Nisbet, R. M. (2007). Fishing the line near marine391
reserves in single and multispecies fisheries. Ecological Applications, 17(4):1039–1054.392
Kuparinen, A. and Merila, J. (2007). Detecting and managing fisheries-induced evolution. Trends393
in Ecology & Evolution, 22(12):652–659.394
Law, R. (2000). Fishing, selection, and phenotypic evolution. ICES Journal of Marine Science,395
57:659–669.396
Law, R. and Grey, D. R. (1989). Evolution of yields from populations with age-specific cropping.397
Evolutionary Ecology, 3:343–359.398
Lester, S. E. and Halpern, B. S. (2008). Biological responses in marine no-take reserves versus399
partially protected areas. Marine Ecology Progress Series, 367:49–56.400
Lester, S. E., Halpern, B. S., Grorrud-Colvert, K., Lubchenco, J., Ruttenberg, B. I., Gaines, S. D.,401
Airame, S., and Warner, R. R. (2009). Biological effects within no-take marine reserves: a global402
synthesis. Marine Ecology Progress Series, 384:33–46.403
Li, E. A. L. (2000). Optimum harvesting with marine reserves. N. Am. J. Fish. Manage., 20:882–404
896.405
Man, A., Law, R., and Polunin, N. V. C. (1995). Role of marine reserves in recruitment to reef406
fisheries: A metapopulation model. Biol. Conserv., 71:197–204.407
McPeek, M. A. and Holt, R. D. (1992). The evolution of dispersal in spatially and temporally408
varying environments. American Naturalist, 140(6):1010–1027.409
Metz, J. (2008). Fitness. In Jørgensen, S. and Fath, F., editors, Encyclopedia of Ecology, pages410
1599–1612. Elsevier.411
Miethe, T., Dytham, C., Dieckmann, U., and Pitchford, J. (2010). Marine reserves and the evolu-412
tionary effects of fishing on size at maturation. ICES Journal of Marine Science, 67(3):412–425.413
Moeller, H. V. and Neubert, M. (2013). Habitat damage, marine reserves, and the value of spatial414
management. Ecological Applications, 23(5):959–971.415
17
Neubert, M. (2003). Marine reserves and optimal harvesting. Ecology Letters, 6:843–849.416
Neubert, M. and Herrera, G. E. (2008). Triple benefits from spatial resource management. Theo-417
retical Ecology, 1:5–12.418
Nowlis, J. S. and Roberts, C. M. (1999). Fisheries benefits and optimal design of marine reserves.419
Fish. Bull., 97:604–616.420
Pala, C. (2014). Tracking fishy behavior, from space. The Atlantic.421
Pezzey, J., Roberts, C. M., and Urdal, B. T. (2000). A simple bioeconomic model of a marine422
reserve. Ecol. Econ., 33:77–91.423
Polacheck, T. (1990). Year around closed areas as a management tool. Nat. Resource Mod., 4:327–424
354.425
Quinn, J. F., Wing, S. R., and Botsford, L. W. (1993). Harvest refugia in marine invertebrate426
fisheries: Models and applications to the red sea urchin, Strongylocentrotus franciscanus. Am.427
Zool., 33:537–550.428
Ratner, S. and Lande, R. (2001). Demographic and evolutionary responses to selective harvesting429
in populations with discrete generations. Ecology, 82(11):3093–3104.430
Sanchirico, J. and Wilen, J. (2001). A bioeconomic model of marine reserve creation. Journal of431
Environmental Economics and Management, 42(3):257–76.432
Sanchirico, J. and Wilen, J. (2005). Optimal spatial management of renewable resources: Matching433
policy scope to ecosystem scale. Journal of Environmental Economics and Management, 50:23–434
46.435
Sanchirico, J. N., Malvadkar, U., Hastings, A., and Wilen, J. E. (2006). When are no-take zones an436
economically optimal fishery management strategy? Ecological Applications, 16(5):1643–1659.437
Schreiber, S. (2012). The evolution of patch selection in stochastic environments. American Natu-438
ralist, 180:17–34.439
Toropova, C., Meliane, I., Laffoley, D., Matthews, E., and Spalding, M., editors (2010). Global440
Ocean Protection: Present Status and Future Possibilities. IUCN.441
18
Tuck, G. and Possingham, H. (1994). Optimal harvesting strategies for a metapopulation. Bulletin442
of Mathematical Biology, 56(1):107–127.443
Uyenoyama, M. and Bengtsson, B. (1982). Towards a genetic theory for the evolution of the sex ratio444
iii. parental and sibling control of brood investment ratio under partial sib-mating. Theoretical445
Population Biology, 22:43–68.446
White, C., Kendall, B. E., Gaines, S., Siegel, D. A., and Costello, C. (2008). Marine reserve effects447
on fisher profit. Ecology Letters, 11:370–379.448
19
Figure 1: Marine reserves (blue polygons) designed to manage scallop harvest off the New EnglandCoast. Dots indicate estimates of fishing effort in 2003, based on satellite tracking of vessels.Warmer colors (green to red) denote more intense activity. The highest intensity of fishing occurredright at MPA borders. Graphic from Fogarty and Murawski (2004).
k2
r2
ε
c2
q2
P1 fished, P2 unfished
P1 fished, P2 in reserve
Both patches fished
P1 in reserve, P2 fished
P1 unfished, P2 fished0.5 1 1.5 2 2 6 10 3 2 1 0.2 0.6
0
0.4
0.2
0.6
0.8
1
1
Figure 2: Optimal fishing effort, in the absence of evolution, in each patch as patch 2 qualityvaries. Patch 2 is the ‘poorer’ patch in every case, with variations in patch 2 parameters noted onthe abscissae. All other parameters are equal between patches, with ki = 10, ri = 2, qi = 1, ci =0.25,m = 4, p = 1. Note that the axis for c2 is flipped, because patch 2 becomes ‘better’ (less costlyto fish) as c2 decreases.
20
π E
1*, E
2* x
, x
*ε
*^
^^
^^
^*
r2 c2k2 q2
* 12
0.5 1 1.5 2 10 24681012 0 0.2 0.4 0.8 10 20 4 6 8 0.6
0
0.5
1
0
1
2
3
0
5
10
0
5
10
0
0.5
1
0
1
2
3
0
5
10
0
5
10
0
0.5
1
0
1
2
3
0
5
10
0
5
10
0
0.5
1
0
1
2
3
0
5
10
0
5
10
Figure 3: ESOHS settlement preference (ε∗), fishing efforts (E∗i ), stock sizes (x∗i ) and sustainablerent (π∗). Parameters not plotted are the same as in Fig. 2. In the middle two rows, the solidcurves indicate effort or stock size in patch 1; the dashed curves depict the same quantities in patch2. Note that the abscissa is reversed when it denotes the value of c2. This makes those figuresconsistent with the rest in that patch 2 becomes either biologically or economically “worse” as onemoves from right to left along the abscissa. Patch 2 is unfished for parameter values to the left ofthe vertical, red, dashed line in each plot.
0 1 20
20
40
60
80
100
r2
2 6 10k2
123c2
0.2 0.6 1q2
No evolution, patch 1 reserveNo evolution, patch 2 reserveEvolution, patch 1 reserveEvolution, patch 2 reserve
% re
nt
Figure 4: Percent of equilibrium rent lost, relative to an optimally managed system with no evolu-tion (in blue) or with evolution (in green). Either patch 1 is in reserve (solid line) or patch 2 is inreserve (dashed line), and effort in the other patch is managed so as to maximize equilibrium rent.Note that when there is no evolution, closing patch 2 may be part of the optimal managementstrategy (when the dashed blue line is at 100%). Parameters are the same as in Fig. 2.
21
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
ϵ
0
0.2
0.4
0.6
0.8
1
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
E1,
E 2
0
0.5
1
1.5
2
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
x 1, x 2
0
1
2
3
4
5
6
t0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
π
0
1
2
3
4
5
6
Figure 5: ESOHS harvesting (dotted lines) versus “reactionary” harvesting (solid lines) in whichthe manager sets effort so as to maximize rent at the current settlement preference (ε) withoutregard to evolutionary stability. Effort and stock size in patch 1 are shown in blue; in patch 2,orange. Mutants (red dots) appear according to a Poisson process with rate µ = 0.01. Each mutantphenotype ε′ is drawn from a normal distribution with mean given by the resident phenotype ε,and standard deviation σ = 0.05, truncated so that 0 < ε′ < 1. Parameters are the same as inFig. 2, except k2 = 1.
22
0 0.5 11
1.1
1.2
q2
0 0.5 10
1
2
3
4
01231
1.1
1.2
01230
1
2
3
4
c2
0 5 101
1.1
1.2
k2
0 5 10
Std.
*
0
1
2
3
4
π
*π
*
π
Figure 6: Ratio of average rent (top row) and standard deviation in rent (bottom row) of the ESOHSstrategy (π∗) compared to “reactionary” harvesting (π∗) in which the manager sets effort so as tomaximize rent at the current settlement preference (ε) without regard to evolutionary stability(cf. Fig. 5). As in earlier figures, all parameters are equal between patches, except that which isnoted on the abscissa. Mutants appear according to a Poisson process at the rate µ = 0.01; theirphenotype is drawn from a normal distribution with mean given by the resident phenotype, andstandard deviation σ = 0.05 (green stars) or σ = 0.5 (black circles), truncated so that 0 < ε′ < 1.Averages were calculated over the time interval [0, 100,000].
23